Transcriptome analysis of damask rose (Rosa damascena Mill.) petals: unveiling the genetic basis of essential oil biosynthesis

Abstract

Background

Identifying the key genes and transcription factors involved in essential oil production in Rosa damascena is an important step toward understanding the molecular mechanisms underlying essential oil biosynthesis. Such knowledge may be applied in breeding programs aimed at improving both the yield and quality of essential oil.

Results

Transcriptomic profiles of two R. damascena genotypes were initially characterized using RNA-Seq followed by de novo transcriptome assembly. De novo transcriptome assembly was conducted using four assemblers. Among these assemblers, Trinity showed the best overall performance, generating the highest number of high-quality unigenes. Comparative transcriptomic analysis revealed distinct expression profiles between the two contrasting genotypes. Gene Ontology (GO) classification indicated that catalytic activity and binding were the predominant categories within the molecular function group. Similarly, metabolic and cellular processes were the most represented categories within biological processes. Key genes involved in monoterpene biosynthesis (NUDX1, HMGCS, ispH, GCPE, GPS) were significantly upregulated in genotype G2, while FDPS, TPS, and GGPS showed higher expression in genotype G1. Additionally, members of the MYB and NAC transcription factor families were markedly upregulated in genotype G2, suggesting their potential involvement in the regulation of essential oil biosynthesis.

Conclusions

This study provides comprehensive insights into the transcriptomic basis of essential oil production in R. damascena. The high essential oil-producing genotype (G2) exhibited enhanced expression of genes associated with the monoterpene biosynthetic pathway, including HMGCS and GPS. MYB and NAC transcription factors were also significantly upregulated in G2. Furthermore, the reduced accumulation of sesquiterpenoids in G2 may be associated with preferential utilization of shared metabolic precursors by the monoterpenoid pathway. These findings identify promising candidate genes and transcription factors for future functional studies and metabolic engineering approaches aimed at improving essential oil yield and quality.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09650-z.

Introduction

Damask rose (Rosa damascena Mill.) is one of the most important aromatic plants for essential oil production, which finds applications in both the perfumery and food industries [1]. The secondary metabolites present in R. damascena petals include essential oil, flavonoids, and anthocyanins, which are of particular importance in breeding programs [2, 3]. The relatively low essential oil content (0.01 to 0.03%) in R. damascena flowers negatively affects the economic value of this crop and consequently reduces producers’ income. The composition of the essential oil, particularly its monoterpene and sesquiterpene constituents, has a significant influence on oil quality and, consequently, its market value [1].

Investigating the petal transcriptome of R. damascena is essential for improving our understanding of the metabolic pathways and key genes involved in the biosynthesis of its bioactive compounds. Such knowledge may facilitate future genetic improvement strategies aimed at enhancing both the yield and quality of essential oil.

Recent advances in sequencing technologies have led to a reduction in sequencing costs [4]. RNA sequencing (RNA-Seq) is a powerful approach for transcriptome analysis, enabling the identification of gene networks and expression patterns associated with secondary metabolite biosynthesis [5]. RNA-Seq enables gene expression profiling and identification of novel transcripts and genes involved in diverse biological processes. Consequently, RNA-Seq has become a valuable tool for transcriptome studies in both model and non-model plant species, particularly those with limited genomic resources [6, 7]. RNA-Seq has been successfully applied to elucidate metabolic pathways in numerous medicinal and aromatic plant species; for instance, it has been used in studies of flavonoid biosynthesis in Thymus daenensis Celak, revealing chalcone isomerase (CHI) and flavonoid 3′-hydroxylase (F3′H) as critical regulators [8]. Additionally, it has contributed to the characterization of terpene synthase transcription factors in Solanum lycopersicum [9], investigation of anthocyanin biosynthesis and phytohormone pathways in Rosa chinensis [10], pigment-related gene identification in Carthamus tinctorius [11], as well as analysis of the ginsenoside biosynthesis pathway in Panax ginseng, including 189 UDP-glycosyltransferase genes [12], and terpenoid pathway profiling in Trachyspermum ammi L [13].

The present study forms part of an ongoing breeding program in R. damascena and builds upon previous investigations of phytochemical diversity among Iranian R. damascena genotypes [14]. This study specifically examines two extreme R. damascena genotypes with contrasting essential oil content and composition. Using comprehensive transcriptome analysis, we aimed to identify candidate genes involved in secondary metabolite biosynthesis and transcription factors (TFs) potentially associated with variation in metabolite accumulation in R. damascena petals.

Materials and methods

Plant material

Petals of two extreme genotypes of R. damascena (G1, G2), characterized by differing phytochemical compounds as indicated in Table 1 [14], were used in this study. Genotype G1 originates from Tehran province and exhibited low essential oil content, whereas genotype G2 originates from Fars province and exhibited higher essential oil content. Petals were collected at the swollen bud stage during early morning hours. Plant materials were obtained from the Horticultural Science Research Station, Faculty of Agriculture, University of Tehran, Karaj, Iran (58°50ʹE and 35°56ʹN, and altitude of 1321 m above sea level). The collected petals were immediately fixed in liquid nitrogen and then stored at -80 °C.

Table 1.

Percentage of phytochemical compositions of two extreme genotypes of Rosa damascena

Genotypes	G1	G2
Essential oil percent	0.015	0.037
Monoterpene hydrocarbons	0.8	3.4
α-Pinene	0.4	2
β-Myrcene	0.2	0.8
Linalool	-	0.7
Oxygenated monoterpenes	5.1	65.7
Phenethyl alcohol	0.4	0.9
Nerol	1.4	9.2
β-Citronellol	1.9	22.3
Neral	0.2	1.0
Geraniol	0.2	26.4
Citronellyl formate	0.3	0.5
Citronellyl acetate	0.3	0.3
Geranyl acetate	0.4	4.3
Sesquiterpene hydrocarbons	1	1.8
β-Bourbonene	-	0.2
β-Elemene	-	0.1
α-Humulene	0.2	0.1
ϒ-Muurolene	0.3	0.3
d -Germacrene	0.2	0.8
Oxygenated sesquiterpenes	7.8	0.1
Elemol	3.3	0.1
Farnesol	0.3	-
Nerolidyl acetate	4.2	-
Aliphatic hydrocarbons	83.1	25.7
n-Heptadecane	1.9	2.4
n-Octadecane	0.4	0.2
Hexadecanal	0.3	-
1-Nonadecene	4.9	3.2
n-Nonadecane	39.5	13.2
n-Heneicosane	24.3	4.1
n-Docosane	1.97	0.1
9-Tricosene	0.4	0.1
n-Tricosane	5.4	0.7
n-Pentacosane	0.4	0.2
Phenylpropanoids	0.5	1.7
Eugenol	-	0.2
Methyl eugenol	0.5	1.5

The original source of this phytochemical data [14] Monoterpene alcohols such as β-Citronellol and geraniol are key contributors to the characteristic fragrance and commercial quality of Rosa damascena essential oil

RNA extraction and sequencing

RNA extraction from the petals G1 and G2 genotypes was performed in two replicates using the CTAB method. For each biological replicate, total RNA was extracted separately from three individual plants and then equal amounts were pooled to form a composite sample. RNA quality was assessed using a NanoDrop 100 (Thermo Fisher Scientific, USA) and Bioanalyzer (Agilent 5400, USA) apparatus. cDNA libraries were constructed using the TruSeq Stranded mRNA LT Sample Prep Kit, followed by sequencing via the Illumina NovaSeq 6000 platform with paired-end 150-bp reads.

Quality control and assembly of reads

The sequencing data were first evaluated using FastQC software, subsequently, Trimmomatic software (version 0.39) was used to trim low-quality reads [15]. De novo transcriptome assembly was performed using four assemblers: Trinity (version 2.10.0), BinPacker-32 (version 32), SPAdes (version 3.14.1), and EvidentialGene (version evigene18may07).

Assessment of assembler quality

To assess the sensitivity, completeness, and accuracy of the assembled transcriptomes, assembly quality was evaluated using rnaQUAST (v2.0.1) and BUSCO, which focuses on the examination of single-copy orthologs [16, 17].

Functional annotation of assembled transcriptomes

The assembled transcriptomes derived from the various assemblers were blasted against SwissProt, Arabidopsis thaliana, Rosa chinensis and Medicinal Plant Genomics Resource (MPGR) protein databases (http://medicinalplantgenomics.msu.edu/species_list.shtml) with E-value.

cutoff of 1e-5.

Gene ontology classification

To assess the Gene Ontology (GO) related to the assembled transcriptome, GO terms were extracted from the UniProt database, and all unigenes were classified using the WEGO software [18].

Identification of functional KEGG pathways

To identify metabolic pathways of R. damascena within the KEGG database, KO numbers for genes associated with these metabolic pathways were initially determined using the KAAS tool. Subsequently, pathway mapping was performed using KEGG Mapper (https://www.genome.jp/kegg/mapper/).

Identification of transcription factors

Transcription factors (TFs) associated with the transcripts were identified through the iTAK tool [19].

Identification and classification of differentially expressed genes

To determine gene expression levels, the frequency of assembled unigenes was estimated using RSEM software [11]. High-quality reads were aligned against the assembled transcript employing Bowtie2 to estimate the number of aligned reads, leading to the calculation of normalized expression values. The analysis of differentially expressed genes (DEGs) between the two genotypes was conducted using edgeR [20]. Lowly expressed transcripts were eliminated before statistical analysis by keeping only those with counts per million (CPM) > 1 in at least two samples. The Trimmed Mean of M-values (TMM) method, which is implemented in edgeR, was used to normalize library size. The estimateDisp function was used to directly estimate biological dispersion from the data. The exactTest function in edgeR was used to measure differential expression between genotypes. Genes with a fold change higher than 2 (|log2FC| > 1) and a false discovery rate (FDR) < 0.01 were deemed differentially expressed. All four libraries (two replicates for G1 and two replicates for G2) were prepared and sequenced simultaneously in a single run on the Illumina NovaSeq 6000 platform, thus eliminating the need for statistical correction of batch effects. Expression levels of GPS, GGPPS, DXS, F3H, MYB, and NAC were validated by qRT-PCR using a QIAGEN. Each 15 µL reaction mixture contained 7.5 µL of SYBR Green Master Mix 2X (ROX), 4.7 µL of water, 2 µL of the cDNA template, and 0.4 µL each of forward and reverse primers (supplementary Table S1). Thermal cycling conditions included an initial denaturation at 95˚C for 15 min, followed by 40 cycles of 15-second denaturation at 95˚C, 20-second annealing at 56–59˚C, and 20-second extension at 72˚C [21]. This served as an independent validation of the RNA-Seq expression data.

Correlation analysis between essential oil compounds, key genes, and TFs

Correlations analyses between major essential oil compounds (Table 1) and the expression levels of key genes of the terpenoid skeleton pathway and important TFs (the five with the highest expression levels) from the MYB, NAC, C3H, and bHLH families were performed using R Studio software.

Results

Quality measurement of RNA extraction

The quality evaluation of the total RNA samples revealed high quality. The RIN (RNA Integrity Number) values for each sample are listed in Table 2. An RIN value above 7 is generally considered suitable for RNA-Seq, and all samples exhibited RIN values exceeding 8.3, confirming their suitability for transcriptome analysis.

Table 2.

Quality control results of extracted RNA from petals of two Rosa damascena genotypes (G1 and G2) in two repetitions (A and B)

No.	Sample Name	Concentration (ng/ul)	Volume (ul)	Total amount(ug)	Integrity value	Sample QC Results
1	G1A	97.000	15.00	1.45500	9.10	Pass
2	G1B	275.000	15.00	4.12500	9.30	Pass
3	G2A	290.000	15.00	4.35000	8.90	Pass
4	G2B	63.000	15.00	0.94500	8.30	Pass

Trimming and quality control of reads

Following sequencing, low-quality reads (Q < 30) and the first 10 bases of each read were trimmed to enhance the quality of the reads. After trimming, the highest number of reads was obtained from sample G1B (32.5 million), while the lowest was from sample G2B (20.3 million) (Table 3).

Table 3.

Checking the number of readings before and after trimming petals of two Rosa damascena genotypes (G1 and G2) in two repetitions (A and B)

Sample Name	The number of readings before trimming	The number of readings after trimming
G1A	22,212,525	21,692,359
G1B	33,266,992	32,579,371
G2A	21,712,747	21,235,194
G2B	21,029,571	20,135,797

Assembly and quality control

Transcriptomes were assembled using four different assemblers. Assembly quality assessment showed that Trinity produced the highest number of unigenes, totaling 274,015, followed by SPAdes with 160,573 unigenes, and BinPacker-32 with the lowest at 110,641. In terms of transcript lengths, the shortest transcript (70 bases) was related to SPAdes, while the longest transcript (17,165 bp) was generated by both EvidentialGene and SPAdes. Trinity also produced the highest total number of assembled bases (244 million), in contrast to EvidentialGene, which recorded the lowest (135 million). The highest number of transcripts containing open reading frames (ORFs) was obtained from Trinity (74,068), while BinPacker-32 yielded the lowest (50,685). Additionally, the reads mapped back to transcripts (RMBT) rate was highest for Trinity at 98.39%, with BinPacker-32 showing the lowest at 94.74% (Table 4). Moreover, BUSCO analysis based on conserved single-copy orthologs indicated that EvidentialGene had the highest number of complete orthologs (250, 98%), whereas BinPacker-32 had the lowest (227, 89%) (Fig. 1). Because no single parameter can fully reflect transcriptome assembly quality, assemblers were chosen based on an overall balance of assembly completeness, transcript recovery, annotation efficiency, and biological utility for subsequent analyses.

Table 4.

Quality control of assemblers using rnaQUAST, Transrate, and RMBT tools

	BinPacker − 32	EvidentialGene	SPAdes	Trinity
n_seqs	110,641	137,022	160,573	274,015
smallest	200	182	70	179
largest	16,114	17,165	17,165	15,905
n_bases	139,573,066	135,262,503	178,626,681	244,657,605
mean_len	1261.49498	987.00107	1110.92387	892.83754
n_under_200	0	113	1284	35
n_over_1k	49,308	46,587	65,467	78,243
n_over_10k	70	38	33	59
n_with_orf	50,685	50,987	51,105	74,068
mean_orf_percent	54.48836	60.40056	47.31298	52.79032
n90	591	417	457	360
n70	1219	917	1232	832
n50	1897	1486	1860	1474
n30	2774	2190	2616	2251
n10	4812	3638	4059	3722
gc	0.4274	0.42465	0.41626	0.41722
bases_n	74	16,004	63,596	0
proportion_n	0	0.00012	0.00036	0
RMBT G1A	95.95	95.32	98.22	98.33
RMBT G1B	96.07	95.37	98.30	98.42
RMBT G2A	93.30	95.01	98.19	98.38
RMBT G2B	93.64	95.09	98.31	98.44
RMBT avr	94.74	95.19	98.25	98.39

Fig. 1.

Comparison of assemblers using single-copy gene orthologs (BUSCO) tools

Although EvidentialGene and BinPacker outperformed Trinity in terms of BUSCO completeness and contiguity-related metrics, Trinity had the highest read-mapping rate, recovered the most ORF-containing transcripts, and generated significantly more annotated sequences than any other reference database. Therefore, Trinity was chosen as the reference assembly for downstream functional analyses.

Functional annotation of unigenes

The assembled transcriptomes were compared against different databases. Across all databases, Trinity consistently yielded the highest number of annotated genes. The highest number of matches was against the Rosa chinensis genome reference, with 194,269 annotated sequences (70.89%), while the lowest number of matches was against the MPGR database with 62,081 genes (23.74%) (Table 5).

Table 5.

Comparison of the number of genes identified in different databases MPGR, SwissProt,

R Ch (Rosa chinensis), and A Th (Arabidopsis thaliana) with E-value cutoff of 1e-5

databases	BinPacker − 32	EvidentialGene	SPAdes	Trinity
MPGR	41,311	31,836	40,649	62,081
SwissProt	62,177	60,821	71,485	110,196
R Ch	88,436	89,508	109,261	194,269
A Th	71,035	67,838	78,328	132,916

Identification and classification of differentially expressed genes

Heatmap analysis of gene expression profiles revealed clear clustering of the two R. damascena genotypes into distinct groups based on their transcriptomic profiles (Fig. 2A). To provide a global overview of differential gene expression, a volcano plot was generated (Fig. 2B). The volcano plot illustrates the distribution of all detected transcripts according to their log2 fold change and false discovery rate (FDR), highlighting significantly upregulated and downregulated genes between the two genotypes. Heatmap analysis of gene expression profiles showed clear clustering of the two genotypes into distinct groups. The numbers of significantly (FDR < 0.01 and log2FC > ± 2) were 7,947 genes for genotype G2 and 7,501 genes for genotype G1. Detailed information for all differentially expressed genes, including log2FC, FDR, normalized expression values, functional annotations, and pathway assignments, is presented in Supplementary Table S1. Furthermore, qRT-PCR validation results were consistent with the RNA-Seq data (Fig. 3, supplementary Table S2).

Fig. 2.

Differential expression analysis of genes in two Rosa damascena genotypes. A Heatmap showing the expression patterns and hierarchical clustering of differentially expressed genes (DEGs) between genotypes G1 and G2. B Volcano plot showing the distribution of all detected transcripts according to log2 fold change and false discovery rate (FDR). Red and blue dots indicate significantly upregulated and downregulated genes, respectively. Differentially expressed genes were identified using the thresholds of FDR < 0.01 and log2 fold change > ± 2

Fig. 3.

Quantitative real-time PCR analysis of relative expression level of GPS, GGPPS, DXS, F3H, MYB, and NAC. Terpenoid and Flavonoid biosynthetic genes of two genotypes (G1 and G2) in Rosa damascene. All genes examined showed expression trends consistent with the RNA-Seq data. Error bars represent the values of the mean ± standard error (SE), with n = 3

Gene ontology classification

The Gene Ontology classification provided a functional categorization of unigenes based on the GO database. The WEGO diagram showed that the transcripts were classified into 57 functional categories (Fig. 4). In the category of molecular function (MF), the highest proportions were attributed to catalytic activity (37.8%) and binding (34.6%). In the biological processes (BP) category, metabolic and cellular processes were the most represented biological process categories at 29.4 and 29.5%, respectively, while in cellular components (CC), the categories of cell, cellular part, and membrane each constituted 23%.

Fig. 4.

GO functional classification of assembled unigenes in Rosa damascene petal. Bars represent the percent and number of assignments of unigenes to each GO term

Enrichment and pathway identification

KAAS annotation assigned KEGG Orthology (KO) terms to 61,877 genes, which were then mapped using the KEGG Mapper tool. The most commonly identified genes were those associated with metabolic pathways (973 genes), biosynthesis of secondary metabolites (476 genes), carbon metabolism (95 genes), amino acid biosynthesis (100 genes), and cofactors biosynthesis (129 genes) (Table 6).

Table 6.

Metabolic pathways of the Rosa damascena petals in the KEGG database

KEGG Pathways	Pathway No.	No. of identified genes	N. of unigenes
Metabolic pathways	01100	973	14,616
Biosynthesis of secondary metabolites	01110	476	8098
Biosynthesis of cofactors	01240	129	1768
Biosynthesis of amino acids	01230	100	1781
Carbon metabolism	01200	95	1877
Terpenoid backbone biosynthesis	900	31	486
Monoterpenoid biosynthesis	902	3	127
Sesquiterpenoid and triterpenoid biosynthesis	909	7	148
Diterpenoid biosynthesis	904	8	56
Carotenoid biosynthesis	906	20	228
Brassinosteroid biosynthesis	905	9	54
Zeatin biosynthesis	908	6	89
Phenylpropanoid biosynthesis	940	15	468
Flavonoid biosynthesis	941	14	187
Flavone and flavonol biosynthesis	944	3	13
Anthocyanin biosynthesis	942	3	13
Isoflavonoid biosynthesis	943	1	31
Isoquinoline alkaloid biosynthesis	950	9	136
Fatty acid biosynthesis	61	17	408

Enrichment analysis of differentially expressed genes demonstrated a pronounced activation of the terpenoid backbone biosynthesis pathway (fold enrichment 1.75-2.00), which forms the foundation for the production of monoterpenes and sesquiterpenes, the major volatile components of essential oil. Concurrently, significant enrichment was observed in phenylpropanoid biosynthesis and its related pathways, particularly through upregulation of aromatic amino acid biosynthesis (phenylalanine, tyrosine) and 2-oxocarboxylic acid metabolism. Genes associated with the biosynthesis of secondary metabolites pathway were significantly enriched suggesting a potential increase in the transcriptional activity of pathways associated with specialized metabolite biosynthesis in the petals. Notably, genes associated with lipid metabolic pathways were significantly enriched (glycerolipid metabolism, fatty acid metabolism, and glycerophospholipid metabolism, with fold enrichments of 1.50–1.75) was also observed (Fig. 5).

Fig. 5.

Parametric analysis of gene set enrichment related to differentially expressed (DE) petal of Rosa damascene. The values of FDR < 0.01 and log2 fold change > ± 2 were used

Analysis of the terpenoid backbone biosynthesis pathway using KEGG revealed that terpenoids in the petals of R. damascena are synthesized via two pathways: mevalonate (MVA) and methylerythritol phosphate (MEP). Expression analysis of genes involved in terpenoid backbone biosynthesis revealed significantly higher expression levels for the genes encoding 2-C-Methyl-D-erythritol 2,4-cyclodiphosphate synthase (ispF), (E)-4-hydroxy-3-methylbut-2-enyl-diphosphate synthase (GCPE), and Geranyl diphosphate synthase (GPS), and 3-Hydroxy-3-Methylglutaryl-CoA Synthase (HMGCS) in genotype G2. In contrast, the expression levels of Farnesyl diphosphate synthase (FDPS) and Geranylgeranyl diphosphate synthase (GGPS) were significantly elevated in genotype G1. Next, the Nudix hydrolase1 (NUDX1) gene, which is responsible for the production of acyclic monoterpenes including geraniol and β-citronellol, was examined and expression of NUDX1 was significantly higher in G2. Also, expression of Sesquiterpene synthase (TPS) genes, which are responsible for the synthesis of sesquiterpenes from farnesyl diphosphate, was significantly higher in G1. (Fig. 6).

Fig. 6.

Expression patterns of the terpenoid skeleton pathway in two genotypes of Rosa damascene petal (G1 and G2) in two repetitions (A and B). The sum of FPKMs of all transcripts related to each gene is used. Expression values are presented as row-scaled z-scores of log2 (FPKM + 1) to highlight relative expression patterns across samples

Transcription factors identified

Our analysis of TFs in R. damascena petals revealed a diverse array of regulatory proteins that orchestrate the biosynthesis of essential oils and other key floral characteristics. A total of 6,219 TFs from 71 families were identified (Fig. 7), with the most abundant families shown in the figure. The most frequently detected TFs belonged to the MYB and C3H families, with counts of 325 and 304, respectively.

Fig. 7.

Percent of identified transcription factor families in the petal of Rosa damascene

Our transcriptomic analysis elucidated distinct expression profiles of TFs between high (G2) and low (G1) essential oil-producing R. damascena genotypes. Analysis of significantly differentially expressed TFs (FDR < 0.01 and log2FC > ± 2) showed marked upregulation of NAC (77.5%) and MYB (60.3%) in G2 compared to G1, while bHLH (37%) and C3H (46%) were significantly downregulated (Fig. 8).

Fig. 8.

Heatmap of differential expression of transcription factor families identified in the petal of two genotypes Rosa damascene (G1 and G2) in two repetitions (A and B). The values of FDR < 0.01 and log2 fold change > ± 2 were used. NAC (A), MYB (B), bHLH (C) and C3H (D)

Correlation analysis

Correlation analysis between essential oil compounds and key biosynthetic genes revealed that monoterpene compounds including Linalool, geraniol, and β-citronellol have a positive correlation (ρ > 0.8) with GPS, HMGCS, NUDX1, and ispH genes. Also, sesquiterpene compounds and linear hydrocarbons, including Elemol, n-Nonadecane, and n-Heneicosane were positively correlated (ρ > 0.8) with TPS, HMGCR, DXR, ispE, ispS, MVD, DXS, and IDI genes. Additionally, significant positive correlations were observed between MYB (MYB16, MYB21, MYB63) and NAC (NAC8, NAC20, and NAC23) and the expression levels of HMGCS and GPS in R. damascena petals (Fig. 9).

Fig. 9.

Correlations between the main essential oil compounds and the expression level of key genes and transcription factors (five transcription factors with the highest expression levels from each of MYBs, NACs, bHLHs and C3Hs). A Correlations between major essential oil metabolites and key terpenoid biosynthesis genes. B Correlations between major essential oil metabolites and transcription factors. C Correlations between key terpenoid biosynthesis genes and transcription factors. Circle color indicates the direction of the correlation (red, positive; blue, negative), circle size represents the correlation coefficient, and asterisks (*) indicate statistically significant correlations (P < 0.05)

Discussion

The choice of assembler has a significant impact on the quality and quantity of the transcriptome [9]. It is essential to evaluate multitude indicators collectively during assembler selection rather than relying on a singular criterion. The BUSCO index highlights that a high number of fragmented orthologs signals assembly issues. Consequently, a greater number of these fragmented orthologs indicates a lower assembly quality [22]. Since various assemblers yield different results across plant species, choosing the appropriate assembler is essential. Although Trinity did not achieve the highest BUSCO completeness score, it outperformed the other assemblers in several metrics, including read-mapping rate, ORF recovery, and annotation success. Therefore, Trinity was selected for analyzing the transcriptome of R. damascena.

According to the enrichment results at the regulatory level, the plant hormone signal transduction pathway (particularly jasmonate and abscisic acid signaling) and the MAPK signaling pathway were significantly enriched. This expression pattern may reflect broader metabolic reprogramming potentially associated with enhanced essential oil biosynthesis; however, direct alterations in carbon flux require validation through metabolomic or fluxomic analyses. These findings suggest that developmental and stress-responsive pathways may be associated with the enhanced oil production phenotype observed in G2 [23, 24]. Interestingly, while these secondary metabolic and regulatory pathways exhibited strong activation, pathways associated with general growth and maintenance, such as ribosome biogenesis and nucleocytoplasmic transport, showed relatively lower enrichment. This observation is consistent with a potential metabolic trade-off between specialized metabolism and primary cellular functions, although further experimental evidence is required to confirm this hypothesis [25]. These findings provide a holistic view of the molecular basis for enhanced essential oil production in R. damascena. The observed expression patterns suggest coordinated regulation of terpenoid and phenylpropanoid biosynthetic pathways, together with changes in precursor-producing pathways and lipid metabolism, that may contribute to differences in essential oil production between the two genotypes. The results also highlight potential targets for metabolic engineering strategies aimed at enhancing both the quantity and quality of R. damascena essential oil. Our combined analysis of gene expression patterns and terpenoid profiles across two R. damascena genotypes suggests the presence of a coordinated metabolic network associated with divergent essential oil production. The high essential oil-producing genotype (G2) exhibited significantly higher expression of several genes involved in both the MVA and MEP pathways, corresponding to its superior essential oil yield (0.037% vs. 0.015% in G1). Particularly noteworthy is the 1.6-fold upregulation of HMGCS in G2, which catalyzes the first committed step of the MVA pathway. Increased HMGCS expression was associated with a higher proportion of oxygenated monoterpenes in G2 (65.7%) compared to G1 (5.1%). This finding is further supported by a 1.8-fold increase in the expression of GPS, the key enzyme responsible for producing geranyl diphosphate. Additionally, NUDX1 expression was increased and TPS expression decreased in G2; these genes are directly responsible for monoterpene and sesquiterpene production, respectively. This expression pattern is consistent with the elevated accumulation of commercially valuable monoterpene alcohols, including β-citronellol and geraniol, in G2.

The metabolite profiles indicated substantial differences in pathway utilization between the two genotypes. While G2 exhibited transcriptional patterns consistent with enhanced monoterpenoid biosynthesis, the low-oil genotype (G1) accumulated substantially more aliphatic hydrocarbons (83.1% vs. 25.7% in G2), which may indicate differences in carbon allocation or metabolic resource utilization associated with wax biosynthesis. This observation is consistent with a potential metabolic trade-off, as G1 accumulated relatively higher levels of oxygenated sesquiterpenes (7.8% vs. 0.1% in G2), consistent with its reduced expression of MVA pathway genes, such as HMGCS. The reduced accumulation of sesquiterpenoids in G2 may be associated with preferential utilization of shared metabolic precursors by the monoterpenoid pathway. These findings are consistent with current models of terpenoid metabolic regulation in aromatic plants, where the coordinated expression of upstream pathway genes and branch-point enzymes dictates the final product profiles. The observed positive correlations between HMGCS, NUDX1, ispH, GPS expression and monoterpenoid accumulation identify these genes as promising candidate regulators that warrant further functional investigation. Furthermore, the upregulation of MYB and NAC transcription factors in G2 was associated with increased expression of several terpenoid biosynthesis genes and may suggest the presence of shared regulatory mechanisms. From an applied perspective, these results highlight specific molecular targets for improving essential oil production.

Overexpression of HMGCS has been previously shown to enhance carbon flux toward terpenoid formation in aromatic plants [26, 27]. increase in GPS expression in G2, together with the higher abundance of geraniol and other oxygenated monoterpenes, is consistent with observations reported in Pelargonium graveolens. Functional characterization of GPS isoforms in P. graveolens has shown that elevated GPS expression directly boosts the availability of geranyl diphosphate and shifts the metabolic flux toward monoterpenes [28]. The marked enrichment of geraniol in G2, together with substantially lower sesquiterpene levels (0.1% vs. 7.8% in G1), suggests pronounced differences in terpenoid accumulation patterns between the two genotypes. These observations raise the possibility that GPS function in R. damascena may differ from that reported in other species. Furthermore, the inverse relationship between monoterpenes and aliphatic hydrocarbons supports models of competing Acetyl-CoA utilization [29]. However, the large difference in aliphatic hydrocarbon abundance between the two genotypes (83.1% vs. 25.7%) may indicate the involvement of additional regulatory mechanisms that remain to be characterized in R. damascena. The coordinated upregulation of HMGCS and GPS together with elevated MYB expression is consistent with previously proposed models of MYB-associated pathway regulation [30]. Nevertheless, the apparent dual activation of both precursor supply and product formation genes may indicate a more integrated regulatory framework, although further experimental validation is required. The petal transcriptome features prominent representation of MYB, bHLH, and AP2/ERF transcription factor families, which are known key regulators of phenylpropanoid and terpenoid biosynthesis pathways [31]. These TF families likely form interconnected networks that control the production of volatile organic compounds, which contribute to the characteristic rose fragrance [32]. Notably, we observed significant representation of WRKY and NAC domain transcription factors, implying a sophisticated regulation of secondary metabolite production in response to developmental and environmental stimuli [33]. Hormone-related TFs, particularly AUX/IAA proteins and bZIP factors, indicate the integration of auxin and ABA signaling pathways in the regulation of petal development and fragrance biosynthesis [34]. Studies on TFs are used to investigate the regulation of specific genes [18]. Research has found that WRKY, MYB, and bHLH transcription factors act as regulators of secondary metabolic pathways [35, 36].

The coordinated upregulation of NAC and MYB in G2 suggests that these transcription factors may participate in a common regulatory network associated with terpenoid biosynthesis, similar to the MYB-NAC complex reported in Arabidopsis thaliana secondary cell wall biosynthesis [37]. The relationship between NAC and MYB in lignin biosynthesis has also been reported in Melilotus albus [38]. This relationship may contribute to the coordinated expression of multiple terpenoid biosynthesis genes. Increased MYB expression was positively associated with higher PAL expression, which may be consistent with enhanced phenylpropanoid pathway activity. This is in accordance with recent metabolomics studies showing higher cinnamate derivatives in high-essential oil genotypes [39]. The NAC expression profile may be associated with biological processes related to senescence regulation in G2, potentially extending the essential oil accumulation window. Our results corroborate the model proposed by Singh et al. [40]. The high expression of MYB in R. damascena petals suggests that this TF could play a key role in regulating secondary metabolites, including terpenoids and flavonoids. This makes MYB and NAC important targets in molecular breeding programs aiming to increase the essential oil content of R. damascena.

According to the present findings, several future research directions can be proposed. These include investigating the regulatory effects of NAC and MYB transcription factors on promoters of key monoterpene biosynthesis genes, such as GPS, using chromatin immunoprecipitation (ChIP) assays. Potential protein–protein interactions and cooperative regulatory mechanisms may also be evaluated using yeast two-hybrid (Y2H) systems. In addition, functional characterization of HMGCS, GPS, MYB, and NAC through gene cloning and transformation approaches could help determine their roles in monoterpene biosynthesis. Silencing of FDPS using RNA interference (RNAi) may further clarify its contribution to sesquiterpene production. Future studies may also investigate the responses of candidate genes and transcription factors to hormone treatments, including jasmonic acid and abscisic acid, as well as abiotic stress conditions. Although the present study identified strong associations between gene expression patterns, transcription factors, and essential oil composition, the results should be interpreted within the limitations of transcriptome-based analyses. RNA-Seq data provide insights into transcriptional regulation but do not directly demonstrate metabolic fluxes, enzyme activities, or causal regulatory interactions. Therefore, the proposed roles of candidate genes and transcription factors should be considered as hypotheses requiring further validation through biochemical, genetic, and functional studies.

Conclusion

Our comprehensive transcriptomic and metabolic analysis of two extreme genotypes of R. damascena provides novel insights into the molecular mechanisms underlying divergent essential oil production. The high essential oil-producing genotype (G2) exhibited a coordinated upregulation of terpenoid biosynthetic genes, including HMGCS, NUDX1, and GPS, alongside transcription factors such as MYB and NAC. This was associated with enhanced expression of genes involved in the MVA pathway and a preferential accumulation of commercially valuable monoterpenes. In addition, this genotype showed lower expression of genes associated with sesquiterpene biosynthesis and reduced levels of aliphatic hydrocarbons, which may indicate differences in metabolic resource allocation between the two genotypes. Enrichment of genes associated with hormone signaling (JA/ABA) and MAPK pathways, together with reduced expression of genes related to growth-associated processes, may be consistent with transcriptional changes linked to specialized metabolism. The identification of candidate genes and transcription factors, particularly HMGCS, GPS, MYB, and NAC, provides a valuable basis for future functional validation and metabolic engineering studies. This study provides a useful transcriptomic resource for investigating the regulation of specialized metabolism in R. damascena and other aromatic crops, and identifies potential molecular targets that may support future breeding and genetic improvement efforts aimed at enhancing essential oil production. Future studies may explore applying these insights to improve breeding strategies for higher essential oil yield in R. damascena.

Authors’ contributions

HA: Conceptualization, Investigation, Data curation, Formal analysis, Writing - original draft. MSH: Supervision, Methodology, Writing, review & editing. RF: Supervision, Methodology, Writing, review & editing. MHM: Data curation, Formal analysis, review & editing. MRN: review & editing. MSHA: Formal analysis.

Funding

Not applicable.

Data availability

The datasets generated and analyzed during the current study are available in the NCBI SRA repository, Data released publicly are available at the following accession numbers: [[SRR35787010](https://dataview.ncbi.nlm.nih.gov/object/SRR35787010), SRR35787011, [SRR35787012] (https://dataview.ncbi.nlm.nih.gov/object/SRR35787011), SRR35787013] (https://www.ncbi.nlm.nih.gov/sra/PRJNA1344159).

Declarations

Ethics approval and consent to participate

All methods performed in this study including the collection of plant materials were in compliance with the relevant institutional, national, and international guidelines and legislation.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.